[SciPy-user] sparse behaviour?

A. M. Archibald peridot.faceted at gmail.com
Mon Oct 9 22:29:32 CDT 2006


On 09/10/06, Stefan van der Walt <stefan at sun.ac.za> wrote:

> Currently, scipy.cov calls numpy.cov, which knows nothing about sparse
> matrices.  I don't think there is any reason why it can't be done,
> it's just that no one has taken the time to implement it.

There are two ways to fix this: just convert the matrix into a dense
matrix and call numpy.cov, or actually implement a covariance-matrix
algorithm for sparse matrices. I would argue *against* doing the
first: if a user is using sparse matrices, they have a reason, whether
because dense matrices would overload memory, or because sparse
algorithms are vastly faster, and secretly exploding their matrix to a
dense one is likely to be confusing and unwanted.

A. M. Archibald


More information about the SciPy-user mailing list